samba - Dec 2011 - File names with unusual characters and linux smbfs clients

I had a weird problem today that brought to light an issue that I
didn't realize was happening on my samba servers. I had a windows user
save a file to a samba share with an "em dash" character in the file
name (em dash is unicode U+2014). The same user tried to find the file
on another computer, a linux client using smbfs to access the same
share and browsing with Nautilus, was unable to find the file.

Doing a little digging, I found that going to a command line and doing
an 'ls' of the directory with the file produced an error message. The
error message was:

ls: cannot access FooFileName?.txt: No such file or directory

Where the "em dash" was replaced by a question mark.

Windows clients can see the file just fine, it looks fine on the
server, and it looks fine when browsing the directory through
nautilus' smbclient, and the ssh/sftp client. It's only when the
directory is mounted via smbfs or cifs that it causes problems.

I am using samba version 3.4.7 (sorry it's not more up to date, it's
on Ubuntu 10.04 LTS).

If anyone can help me figure out why this is possibly happening, or
how to point me towards a solution. I would really appreciate it.
Thanks.

From: Stephan Stachurski <ses1984 at gmail.com>
Date: Fri, 2 Dec 2011 16:32:17 -0500
> Windows clients can see the file just fine, it looks fine on the
> server, and it looks fine when browsing the directory through
> nautilus' smbclient, and the ssh/sftp client. It's only when the
> directory is mounted via smbfs or cifs that it causes problems.
Have you set iocharset (and codepage if smbfs) properly?

Both smbfs and cifs are not part of Samba, they are part of Linux
kernel.

Does the same problem occur when you you access an "U+2014" filename
on Windows share via smbfs/cifs?

---
TAKAHASHI Motonobu <monyo at samba.gr.jp>